Tensorflow-custom-object-detection-with-ip-camera

The project developed using TensorFlow object detection api to detect unauthorized objects(in this case mobiles) indoors using security cameras/ ip camera/ CCTV cameras... It works in both normal mode & in night vision mode.

A faster R-CNN pretrained model was used with custom dataset to train the detection model "rnn".

Python multi-threading library was used to optimize the rstp stream from the ip camera. A HIK Vision ip camera was used.

Installation guides

• Use anaconda.
  

• install tensorflow using : conda install tensorflow-gpu
  

• It will take care of all the necessary drivers to run tensorflow-gpu ...
  

• clone this repository.
  

• install dependencies of tensorflow. follow this guide : TF_INstall
  

• go to Tensorflow-custom-object-detection-with-ip-camera\research & run:
  protoc object_detection/protos/*.proto --python_out=.
  

• then run: python setup.py build
  

• & then : python setup.py install
  

• object detection module will be ready to be deployed.

you need to install opencv & imutils library too
pip install imutils
pip install opencv-contrib-python

go to: Tensorflow-custom-object-detection-with-ip-camera\research\object_detection
open ip_camera.py & edit it to give your own ipcamera's rstp address.

run : python od_ip.py
you should be good to go.

Application

• Piracy prevention in cinema halls.
  
• Places where recording devices(mobile) are not allowed.